# CUDA artifact selection for Pascal GPU

**URL:** <https://discourse.julialang.org/t/cuda-artifact-selection-for-pascal-gpu/137434>\
**Category:** GPU\
**Created:** [June 3, 2026, 1:23pm UTC](https://discourse.julialang.org/t/cuda-artifact-selection-for-pascal-gpu/137434 "2026-06-03T13:23:21Z")\
**Posts on this page:** 6\
**Page:** 1

<div class="post-metadata">

**Author:** ![EGau](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/egau/32/221186_2.png) [@EGau](https://discourse.julialang.org/u/EGau)\
**Post date:** [June 3, 2026, 1:23pm UTC](https://discourse.julialang.org/t/cuda-artifact-selection-for-pascal-gpu/137434/1 "2026-06-03T13:23:21Z")

</div>

I am using CUDA.jl on a Pascal (sm\_61) hardware.

Till now having driver 570xx (575xx also) worked when restricting cuDNN to v1.4.4 (so nvidia cudnn is \<v9.10.0, later Pascal is deprecated). This restricted also CUDA.jl to version 5.8.5. Tying runtime\_version to v12.x downloaded the right artifact and it worked.

Now I tried to use the latest driver version to support Pascal, 580.xx, and did the same restrictions as above. This time it crashed already using a minimal example with CuArray. dmesg showed me a segfault in libcuda.so.595.58.03 which points to a driver/artifact-library mismatch.

So my question: is it possible to get an artifact having v580 so this mismatch cannot happen?

> julia\> CUDA.versioninfo()  
> CUDA toolchain:
> 
> - runtime 12.4, artifact installation
> - driver 580.159.4 for 13.2
> - compiler 12.9
> 
> CUDA libraries:
> 
> - CUBLAS: 12.4.5
> - CURAND: 10.3.5
> - CUFFT: 11.2.1
> - CUSOLVER: 11.6.1
> - CUSPARSE: 12.3.1
> - CUPTI: 2024.1.1 (API 12.4.0)
> - NVML: 13.0.0+580.159.4
> 
> Julia packages:
> 
> - CUDA: 5.8.5
> - CUDA\_Driver\_jll: 13.2.1+0
> - CUDA\_Compiler\_jll: 0.2.2+0
> - CUDA\_Runtime\_jll: 0.19.2+0
> 
> Toolchain:
> 
> - Julia: 1.12.6
> - LLVM: 18.1.7
> 
> Preferences:
> 
> - CUDA\_Runtime\_jll.version: 12.4
> 
> 1 device:  
> 0: Quadro P2000 (sm\_61, 3.682 GiB / 5.000 GiB available)

---

<div class="post-metadata">

**Author:** ![maleadt](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/maleadt/32/10097_2.png) [@maleadt](https://discourse.julialang.org/u/maleadt)\
**Post date:** [June 4, 2026, 11:31am UTC](https://discourse.julialang.org/t/cuda-artifact-selection-for-pascal-gpu/137434/2 "2026-06-04T11:31:07Z")

</div>

> [@EGau](#):
>
> This time it crashed already using a minimal example with CuArray. dmesg showed me a segfault in libcuda.so.595.58.03 which points to a driver/artifact-library mismatch.

That is probably a bug in the CUDA forwards compatible driver. Try running with `JULIA_CUDA_USE_COMPAT=false`.

---

<div class="post-metadata">

**Author:** ![Stimpleton](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/stimpleton/32/222898_2.png) [@Stimpleton](https://discourse.julialang.org/u/Stimpleton)\
**Post date:** [July 27, 2026, 6:45pm UTC](https://discourse.julialang.org/t/cuda-artifact-selection-for-pascal-gpu/137434/3 "2026-07-27T18:45:20Z")

</div>

Hi, I seem to run into a similar issue with a Pascal P1000 GPU.  
I am trying to use CUDA in combination with WaterLily but get the following error:

CUDA error: operation not supported (code 801, ERROR\_NOT\_SUPPORTED)

And the error gets thrown by this line of code:

GPUsim = sphere(3\*2^5,2^6;T=Float32,mem=CuArray);

I have fresh installed the latest driver, but this does not fix the issue.

As suggested in this thead I used : JULIA\_CUDA\_USE\_COMPAT: false  
But this did not make it work.

I have also tried to use an earlier version of CUDA, version 12.8, by using:  
CUDA.set\_runtime\_version!(v"12.8")

This also has not resolved the issue.

I have the following version info, What strikes me is that my driver is ‘unknown’. What could be my issue?

julia\> CUDA.versioninfo()  
CUDA toolchain:

- runtime 12.8.0, artifact installation
- unknown driver for 13.0
- compiler 12.9.41, artifact installation

CUDA libraries:

- cuBLAS: 12.8.4
- cuSPARSE: 12.5.8
- cuSOLVER: 11.7.3
- cuFFT: 11.3.3
- cuRAND: 10.3.9
- CUPTI: 2025.1.1 (API 12.8.0)
- NVML: missing

Julia packages:

- CUDACore: 6.2.1
- GPUArrays: 11.5.8
- GPUCompiler: 1.23.0
- KernelAbstractions: 0.9.42
- CUDA\_Driver\_jll: 13.3.0+1
- CUDA\_Compiler\_jll: 0.4.4+1
- CUDA\_Runtime\_jll: 0.23.0+1
- NVPTX\_LLVM\_Backend\_jll: 22.1.7+1

Toolchain:

- Julia: 1.12.6
- LLVM: 18.1.7

Environment:

- JULIA\_CUDA\_USE\_COMPAT: false

Preferences:

- CUDA\_Runtime\_jll.version: 12.8

1 device:  
0: Quadro P1000 (sm\_61, 3.285 GiB / 4.000 GiB available)  
compiles to sm\_61 / PTX 8.8

 ![image](https://global.discourse-cdn.com/julialang/original/3X/f/e/fe682294624eea23f66ff386af5aeff451676ce6.png)

---

<div class="post-metadata">

**Author:** ![maleadt](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/maleadt/32/10097_2.png) [@maleadt](https://discourse.julialang.org/u/maleadt)\
**Post date:** [July 28, 2026, 8:49am UTC](https://discourse.julialang.org/t/cuda-artifact-selection-for-pascal-gpu/137434/4 "2026-07-28T08:49:09Z")

</div>

> [@Stimpleton](#):
>
> CUDA error: operation not supported (code 801, ERROR\_NOT\_SUPPORTED)

Please file an issue including a backtrace and MWE. Or check [NVIDIA GeForce GTX 1080 GPU (compute capability 6.1) is not fully supported by CUDA 12.9.0. · Issue #3187 · JuliaGPU/CUDA.jl · GitHub](https://github.com/JuliaGPU/CUDA.jl/issues/3187) if the issue matches.

FWIW, the `JULIA_CUDA_USE_COMPAT` workaround should never be needed. The JLL should detect problematic configurations now, as of [CUDA\_Driver: gate compat driver on device capability - Pull Request #14178 - JuliaPackaging/Yggdrasil - GitHub](https://github.com/JuliaPackaging/Yggdrasil/pull/14178)

---

<div class="post-metadata">

**Author:** ![Stimpleton](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/stimpleton/32/222898_2.png) [@Stimpleton](https://discourse.julialang.org/u/Stimpleton)\
**Post date:** [July 28, 2026, 12:10pm UTC](https://discourse.julialang.org/t/cuda-artifact-selection-for-pascal-gpu/137434/5 "2026-07-28T12:10:58Z")

</div>

Hi,  
I have now tried to test CUDA on my desktop PC. This uses a GTX 1060 GPU (instead of Quadro P1000) but still CUDA compute capability 6.1

I have made a MWE where I run two basic CUDA tests and pass them. But the third test fails at the same line as before and gives the same code 801 error:

```julia-auto
using CUDA
CUDA.versioninfo()

@assert CUDA.functional()

#simple CUDA test (RESULT: TEST PASSED)
using Test
N = 2^20
x_d = CUDA.fill(1.0f0, N) # a vector stored on the GPU filled with 1.0 (Float32)
y_d = CUDA.fill(2.0f0, N) # a vector stored on the GPU filled with 2.0

y_d .+= x_d
@test all(Array(y_d) .== 3.0f0)

#GPU kernel test (RESULT: TEST PASSED)
function gpu_add1!(y, x)
    for i = 1:length(y)
        @inbounds y[i] += x[i]
    end
    return nothing
end

fill!(y_d, 2)
@cuda gpu_add1!(y_d, x_d)
@test all(Array(y_d) .== 3.0f0)

#Test WaterLily package in combination with CUDA
using WaterLily

function sphere(n, m; Re=100, U=1, T=Float64, mem=Array)
    radius, center = m/8, m/2 - 1
    body = AutoBody((x, t) -> √sum(abs2, x .- center) - radius)
    Simulation((n, m, m), (U, 0, 0), 2radius; ν=U*2radius/Re, body, T, mem)
end

GPUsim = sphere(3*2^5, 2^6; T=Float32, mem=CuArray) # <-- error thrown here

```

The error gives the following stacktrace:

```julia-auto
ERROR: CUDA error: operation not supported (code 801, ERROR_NOT_SUPPORTED)
Stacktrace:
  [1] checked_cuModuleLoadDataEx(_module::Base.RefValue{…}, image::Ptr{…}, numOptions::Int64, options::Vector{…}, optionValues::Vector{…})
    @ CUDACore C:\Users\Gebruiker\.julia\packages\CUDACore\NlVPI\lib\cudadrv\module.jl:27
  [2] CuModule(data::Vector{UInt8}, options::Dict{CUDACore.CUjit_option_enum, Any})
    @ CUDACore C:\Users\Gebruiker\.julia\packages\CUDACore\NlVPI\lib\cudadrv\module.jl:59
  [3] CuModule
    @ C:\Users\Gebruiker\.julia\packages\CUDACore\NlVPI\lib\cudadrv\module.jl:48 [inlined]
  [4] link(job::GPUCompiler.CompilerJob, compiled::@NamedTuple{image::Vector{UInt8}, entry::String})
    @ CUDACore C:\Users\Gebruiker\.julia\packages\CUDACore\NlVPI\src\compiler\compilation.jl:504
  [5] actual_compilation(cache::Dict{…}, src::Core.MethodInstance, world::UInt64, cfg::GPUCompiler.CompilerConfig{…}, compiler::typeof(CUDACore.compile), linker::typeof(CUDACore.link))
    @ GPUCompiler C:\Users\Gebruiker\.julia\packages\GPUCompiler\BSi1T\src\execution.jl:270
  [6] cached_compilation(cache::Dict{…}, src::Core.MethodInstance, cfg::GPUCompiler.CompilerConfig{…}, compiler::Function, linker::Function)
    @ GPUCompiler C:\Users\Gebruiker\.julia\packages\GPUCompiler\BSi1T\src\execution.jl:159
  [7] macro expansion
    @ C:\Users\Gebruiker\.julia\packages\CUDACore\NlVPI\src\compiler\execution.jl:456 [inlined]
  [8] macro expansion
    @ .\lock.jl:376 [inlined]
  [9] cufunction(f::WaterLily.var"#gpu_##kern_#546#applyV!##1"{…}, tt::Type{…}; kwargs::@Kwargs{…})
    @ CUDACore C:\Users\Gebruiker\.julia\packages\CUDACore\NlVPI\src\compiler\execution.jl:451
 [10] cufunction
    @ C:\Users\Gebruiker\.julia\packages\CUDACore\NlVPI\src\compiler\execution.jl:448 [inlined]
 [11] #kernel_compile#753
    @ C:\Users\Gebruiker\.julia\packages\CUDACore\NlVPI\src\compiler\execution.jl:60 [inlined]
 [12] macro expansion
    @ C:\Users\Gebruiker\.julia\packages\CUDACore\NlVPI\src\compiler\execution.jl:183 [inlined]
 [13] (::KernelAbstractions.Kernel{…})(::CuArray{…}, ::Vararg{…}; ndrange::NTuple{…}, workgroupsize::Nothing)
    @ CUDACore.CUDAKernels C:\Users\Gebruiker\.julia\packages\CUDACore\NlVPI\src\CUDAKernels.jl:125
 [14] (::WaterLily.var"##kern#545#applyV!##4")(c::CuArray{Float32, 4, CUDACore.DeviceMemory})
    @ WaterLily C:\Users\Gebruiker\.julia\packages\WaterLily\yOkji\src\core.jl:142
 [15] macro expansion
    @ C:\Users\Gebruiker\.julia\packages\WaterLily\yOkji\src\core.jl:144 [inlined]
 [16] applyV!
    @ C:\Users\Gebruiker\.julia\packages\WaterLily\yOkji\src\Flow.jl:82 [inlined]
 [17] apply!(f::Function, c::CuArray{Float32, 4, CUDACore.DeviceMemory})
    @ WaterLily C:\Users\Gebruiker\.julia\packages\WaterLily\yOkji\src\Flow.jl:81
 [18] Flow(N::Tuple{…}, uBC::Tuple{…}; mem::Type, Δt::Float64, ν::Float64, g::Nothing, u0::Nothing, uλ::Nothing, perdir::Tuple{}, exitBC::Bool, λ::typeof(quick), T::Type)
    @ WaterLily C:\Users\Gebruiker\.julia\packages\WaterLily\yOkji\src\Flow.jl:140
 [19] Flow
    @ C:\Users\Gebruiker\.julia\packages\WaterLily\yOkji\src\Flow.jl:133 [inlined]
 [20] #242
    @ C:\Users\Gebruiker\.julia\packages\WaterLily\yOkji\src\WaterLily.jl:96 [inlined]
 [21] Simulation(dims::Tuple{…}, uBC::Tuple{…}, L::Float64; Δt::Float64, ν::Float64, g::Nothing, U::Nothing, ϵ::Int64, perdir::Tuple{}, u0::Nothing, uλ::Nothing, exitBC::Bool, λ::typeof(quick), body::AutoBody{…}, flow_ctor::WaterLily.var"#240#241"{…}, pois_ctor::WaterLily.var"#244#245"{…}, T::Type, mem::Type)
    @ WaterLily C:\Users\Gebruiker\.julia\packages\WaterLily\yOkji\src\WaterLily.jl:103
 [22] Simulation
    @ C:\Users\Gebruiker\.julia\packages\WaterLily\yOkji\src\WaterLily.jl:93 [inlined]
 [23] #sphere#19
    @ d:\Julia_VisualStudio\Error801_Reproduce_Folder\Error_801.jl:34 [inlined]
 [24] top-level scope
    @ d:\Julia_VisualStudio\Error801_Reproduce_Folder\Error_801.jl:37
Some type information was truncated. Use `show(err)` to see complete types.

```

 ![image](https://global.discourse-cdn.com/julialang/original/3X/3/1/31bba5a4390f1b6917b515ba279e23fdff8b10e0.png)

I have the following info on CUDA and Julia:

```julia-auto
CUDA toolchain: 
- runtime 12.9.0, artifact installation
- unknown driver for 13.0
- compiler 12.9.41, artifact installation

CUDA libraries: 
- cuBLAS: 12.9.1
- cuSPARSE: 12.5.10
- cuSOLVER: 11.7.5
- cuFFT: 11.4.1
- cuRAND: 10.3.10
- CUPTI: 2025.2.1 (API 12.9.1)
- NVML: missing

Julia packages: 
- CUDACore: 6.2.1
- GPUArrays: 11.5.8
- GPUCompiler: 1.23.0
- KernelAbstractions: 0.9.42
- CUDA_Driver_jll: 13.3.0+1
- CUDA_Compiler_jll: 0.4.4+1
- CUDA_Runtime_jll: 0.23.0+1
- NVPTX_LLVM_Backend_jll: 22.1.7+1

Toolchain:
- Julia: 1.12.6
- LLVM: 18.1.7

1 device:
  0: NVIDIA GeForce GTX 1060 6GB (sm_61, 5.075 GiB / 6.000 GiB available)
     compiles to sm_61 / PTX 8.8

```

 ![image](https://global.discourse-cdn.com/julialang/original/3X/8/0/80b08f0276a1982fd137d0e2c2fc28bad194b782.png)

---

<div class="post-metadata">

**Author:** ![maleadt](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/maleadt/32/10097_2.png) [@maleadt](https://discourse.julialang.org/u/maleadt)\
**Post date:** [July 29, 2026, 7:49am UTC](https://discourse.julialang.org/t/cuda-artifact-selection-for-pascal-gpu/137434/6 "2026-07-29T07:49:28Z")

</div>

Please refer to [NVIDIA GeForce GTX 1080 GPU (compute capability 6.1) is not fully supported by CUDA 12.9.0. · Issue #3187 · JuliaGPU/CUDA.jl · GitHub](https://github.com/JuliaGPU/CUDA.jl/issues/3187#issuecomment-5114684350)